Research On Some Problems About Two Sorts Of Uncertain Problem In IR

Studies on uncertain problems is very significant to improve the performance and efficiency of IRS,especially for heightening the rate of precision parameter, which people generally care about. It is a key problem for the construction and implement of the future Web network. It is also author's motivation for the study on some problems about two sorts of uncertain problem in IRS.The background of this thesis is "Development and Application of Electronic Archives Pigeonhole System in the Environment of Integrated Network" supported by a key research fund of the National Archives Bureau, and an uncertain problem is proposed and studied when the performance and efficiency of electronic archives system is discussed. There are two sorts of uncertain problems focused by pursuers, one denotes the uncertain semantic knowledge of a query or a document topic; another indicates the relevance between query model and document model. When two sorts of uncertain problems are discussed, various theoretical models including Ontology model, CBR Model (Cased-Based Reason Model), BCN Model, Risk Analysis Model and Interactive Computing Model, are used and explored. These models have their characteristics, and preconditions considered for analyzing and disposing two sorts of uncertain problems. In this thesis, we first briefly introduce the background, intention, significance and current condition of this research; detailedly enumerate some basic theoretical models and some main research methods for disposing uncertain problems in IRS. In this thesis, two sorts of uncertain problem in IRS are deeply studied from various theoretical models, and some innovational views are presented as follow:When two sorts of uncertain problem are studied, the evaluation issue about uncertainty in IRS is becoming important and appears from A to izzard in the thesis. For semantic uncertainty in IRS, three relevant computing strategies presented by Ehrig and Maedche is discussed, also applied in the evaluation of a document, and the evaluation issue of these parameters, for example, Recall, Precision and P_R, based on three relevant computing strategies are proposed and compared. Under the condition of explicit concepts provided by a user, we can draw a conclusion that it is of significant method for ontology applied in IRS to reduce the influence of uncertainty, and to improve the performance and efficiency of IRS. BCN Model is also a good tool for disposing uncertainty in IRS. The network structure of BCN is viewed as a kind of probabilistic distributing modality of knowledge based on whole document space. To find optimization structure, or approximate optimization structure becomes key problem for decreasing the level of uncertainty. Also, under the condition of explicit concepts provided by a searcher, Comparing with Ontology model, BCN model extends the semantic concept of query through quantitative means, such as the form of probability, and a searcher can obtain the description of evaluation for related documents in the collections. BCN has two models, SBCN and ABCN. Based on a searcher's profile, ABCN_UP is proposed and built when referring to ABCN. It shows the correlative relations between concept and concept, or between concept and document in the interest space of a special searcher. ABCN_UP has its advantages for decreasing the searching space of mutual information, and satisfying subjective information needs of a special searcher. When two sorts of uncertain problem are discussed, also one aspect of accessing behavior of a user should not be ignored. It is inevitably connected with correlative uncertainty between query model and document model. In the thesis, from the view of effective expression of knowledge for accessing behavior, a profile CBR model based on an interests space of a special user is proposed and built, and at the same time a searching method for the model is also presented accordingly. In addition, Risk Analysis Model is also an appreciable method for synthetically disposing two sorts of uncertainty in...